1. Using Corpus Linguistics Tools to Aid Research in the Social Sciences
Mike Scott
Aston University
Friedrich-Alexander University Erlangen, 25 January 2016

3. Bootyful, cyw, scrims
Bootyful, cyw, scrims. Do you know these words? If not, you soon might. They are some of the fastest growing words from online niches around the world, as identified by new software that charts the rise of language online.
Bootyful, an alternative spelling for beautiful, has had a dramatic rise in usage on Twitter in South Wales. Cyw (coming your way) has become popular in the north of the country. Scrims comes from gaming forums, where it refers to practice sessions before competitive games.
The software that found these words was developed by Daniel Kershaw and his supervisor, Matthew Rowe, at Lancaster University, UK. Kershaw and Rowe took established methods lexicographers use to chart the popularity of words, translated them into algorithms, then applied them to 22 million words worth of twitter and Reddit posts.
Their goal is to peer into the niche portions of the internet, and chart novel language making its foray out into the mainstream. “If we see an innovation taking off on Reddit or Twitter, the question is what point is it going to appear in a newspaper,” says Rowe.
Kershaw and Rowe’s algorithms don’t just pick out frequently used words, but words that have gone through a sudden rise in popularity. This comes with some complications. The five fastest rising words in central London for the period they studied were all Spanish or Portuguese, unlikely to be reflecting the reality of London’s language scene. (

5. Agenda
Why Corpus Tools?
Which Corpus Tools?
How?

6. My reason for interest language teaching English for Academic Purposes
Latin America
students struggling to understand main points
but getting bogged down in detail

7. Social Science Research Agenda
Explore events
Understand causes
Understand processes

12. Research Agenda
Make sense of complexity

13. Kintsch & van Dijk (1970s) macro-rules
Generalization: use “super-propositions”
Deletion: omit unwanted detail
Construction: use entailment to draw inferences

14. A Text Linguistic Objective of the 1970s
To come up with a set of macro-propositions which could represent the gist of the text.

15. Kinstch & van Dijk macro-rules
Generalization: use “super-propositions”
Deletion: omit unwanted detail
Construction: use entailment to draw inferences

16. Generalization
“Of a sequence of propositions we may substitute any subsequence by a proposition defining the immediate superconcept of the micropropositions”
Mary was drawing a picture. Sally was jumping rope and Daniel was building something with Lego blocks.
The children were playing.

17. Deletion
“Of a sequence of propositions we may delete all those denoting an accidental property of a discourse referent”
A girl in a yellow dress passed by.
1. A girl passed by. (2. She was wearing a dress. 3. The dress was yellow.)

18. Construction
“Of a sequence of propositions we may substitute each subsequence by a proposition if they denote normal conditions, components or consequences of the macroproposition substituting them.” 
John went to the station. He bought a ticket, started running when he saw what time it was and was forced to conclude that his watch was wrong when he reached the platform.
John missed the train.

19. Corpus Linguistics
but now in 2016?

20. Standard corpora British National Corpus
Corpus of Contemporary American English (COCA)

22. Your own corpus
In contrast to other well-known corpora and corpus archives (such as the British National Corpus), however, the German Reference Corpus is explicitly not designed as a balanced corpus: The distribution of DeReKo texts across time or text types does not match some predefined percentages.
This conception complies with the fact that whether or not a given corpus constitutes a balanced or even representative language sample may only be assessed with respect to a specific language domain (i.e., the statistical population). Because different linguistic investigations generally aim at different language domains, the declared purpose of the German Reference Corpus is to serve as a versatile superordinate sample, or primordial sample (German: Ur-Stichprobe) of contemporary written German, from which corpus users may draw a specialised subsample (a so-called virtual corpus) to represent the language domain they wish to investigate.
( Jan 2016)

23. Your own texts From your students Your research archive
LexisNexis and other standard text collections
Project Gutenberg
Oxford Text Archive

24. text patterns

25. Example of corpus

26. Twitter

27. txtLAB450

28. accompanying spreadsheet

29. Limitations Corpus tools typically ignore images numbers, dates
equations
variations in typeface
hyperlinks
related sound or video files

30. Problems Multiple formats PDF format One text = one file?
Incomplete texts
Duplicated texts

31. Formats ASCII, ANSI (one byte per character)
legacy formats from 1960s to 1990s (DOS, Windows, Mac, IBM etc.)
UTF8 varied bytes per character
UTF16 allows for 65,000 characters, fixed 2-bytes per character

32. PDF
save as?

33. converting PDF to plain text….
Adobe Reader
Save As…
Export as Word .docx

35. Adobe Acrobat
OCR
Save as .doc

36. Why Corpus Tools process larger amounts of text
transform the text in varied ways
seeking patterns

37. Which Corpus Tools?
online corpus tools
stand-alone
grammar patterns

38. How simple word lists concordances collocation patterns
dispersion plots

39. at a basic level… let you see the overall vocabulary
multiple examples of words & phrases
can be broken down by
number of texts
location in text
context words

40. dealing with large amounts of data
sorting
filtering out

41. corpus-based or corpus-driven
corpus-based: uses a corpus to try to find examples of something where the underlying research categories already exist
looking through a large text corpus for references to austerity and seeking collocates

42. corpus-based or corpus-driven
corpus-driven: explores a corpus trying to find out what is identified as typical or outstanding
looking through a large text corpus concerning austerity, seeking typical key words

43. Issues What is the unit of text we are working with?
Can you see the text patterns?

44. Levels of Context What is the unit of text we are working with?
single words
n-grams
paragraphs
whole texts
genres

45. Corpus Linguistics can operate at any of these levels
and can use comparison

46. Choice of reference corpus

47. Finding patterns

49. Conclusions Corpus Linguistics tools: relevant to the Social Scientist
traditional tools to be retained
limitations
do not give answers but pointers